1、IntroductionWelcomeMachine LearningAndrew NgAndrew NgSPAMAndrew NgMachine Learning- Grew out of work in AI- New capability for computers Examples:- Database miningLarge datasets from growth of automation/web.E.g., Web click data, medical records, biology, engineering- Applications cant program by ha

2、nd.E.g., Autonomous helicopter, handwriting recognition, most of Natural Language Processing (NLP), Computer Vision.Andrew Ngew capability for computersles:atabase miningLarge datasets from growth of automation/web.E.g., Web click data, medical records, biology, engineeripplications cant program by

3、hand.E.g., Autonomous helicopter, handwriting recognition, Natural Language Processing (NLP), Computer Vision.Machine Learning- Grew out of work in AI- NExamp- Dng- Amost ofAndrew NgMachine Learning- Grew out of work in AI- New capability for computers Examples:- Database miningLarge datasets from g

4、rowth of automation/web.E.g., Web click data, medical records, biology, engineering- Applications cant program by hand.E.g., Autonomous helicopter, handwriting recognition, most of Natural Language Processing (NLP), Computer Vision.Andrew NgMachine Learning- Grew out of work in AI- New capability fo

5、r computers Examples:- Database miningLarge datasets from growth of automation/web.E.g., Web click data, medical records, biology, engineering- Applications cant program by hand.E.g., Autonomous helicopter, handwriting recognition, most of Natural Language Processing (NLP), Computer Vision.- Self-cu

6、stomizing programsE.g., Amazon, Netflix product recommendationsAndrew NgMachine Learning- Grew out of work in AI- New capability for computers Examples:- Database miningLarge datasets from growth of automation/web.E.g., Web click data, medical records, biology, engineering- Applications cant program

7、 by hand.E.g., Autonomous helicopter, handwriting recognition, most of Natural Language Processing (NLP), Computer Vision.- Self-customizing programsE.g., Amazon, Netflix product recommendations- Understanding human learning (brain, real AI).Andrew NgAndrew NgAndrew NgIntroductionWhat is machinelear

8、ningMachine LearningMachine Learning definitionAndrew NgMachine Learning definitionArthur Samuel (1959). Machine Learning: Field of study that gives computers the ability to learn without being explicitly programmed.Andrew NgMachine Learning definitionArthur Samuel (1959). Machine Learning: Field of

9、 study that gives computers the ability to learn without being explicitly programmed.Andrew NgMachine Learning definitionArthur Samuel (1959). Machine Learning: Field of study that gives computers the ability to learn without being explicitly programmed.Tom Mitchell (1998) Well-posed Learning Proble

10、m: A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E.Andrew Ng“A computer program is said to learn from experience E with respect tosome task T and some performance m

11、easure P, if its performance on T, as measured by P, improves with experience E.”Suppose your email program watches which emails you do or do not mark as spam, and based on that learns how to better filterspam.What is the task T in this setting?Classifying emails as spam or not spam.Watching you lab

12、el emails as spam or not spam.The number (or fraction) of emails correctly classified as spam/not spam.None of the abovethis is not a machine learning problem.me task T and some performance measure P, if itsperformances measured by P, improves with experience E.”ose your email program watches which

13、emails you do ark as spam, and based on that learns how to better.What is the task T in this setting?Classifying emails as spam or not spam. Watching you label emails as spam or not spam.The number (or fraction) of emails correctly classified as spam/not sNone of the abovethis is not a machine learn

14、ing problem.“A computer program is said to learn from experience E with respect tosoaSupp not m spamon T,or dofilterpam.“A computer program is said to learn from experience E with respect tosome task T and some performance measure P, if its performance on T, as measured by P, improves with experienc

15、e E.”Suppose your email program watches which emails you do or do not mark as spam, and based on that learns how to better filterspam.What is the task T in this setting?Classifying emails as spam or not spam.Watching you label emails as spam or not spam.The number (or fraction) of emails correctly c

16、lassified as spam/not spam.None of the abovethis is not a machine learning problem.Machine learning algorithms:-Supervised learningUnsupervised learningOthers: Reinforcement learning, recommendersystems.Also talk about: Practical advice for applyinglearning algorithms.Andrew NgAndrew NgAndrew NgIntr

17、oductionSupervisedLearningMachine LearningHousing price prediction.400300Price ($)in 1000s200100005001000150020002500Size in feet2Supervised Learning“right answers” givenRegression: Predict continuousvalued output (price)Andrew NgBreast cancer (malignant, benign)ClassificationDiscrete valued output

18、(0 or 1)1(Y)Malignant?0(N)Tumor SizeTumor SizeAndrew Ng-Clump ThicknessUniformity of Cell SizeUniformity of Cell Shape-AgeTumor SizeAndrew NgYoure running a company, and you want to develop learning algorithms to address eachof two problems.Problem 1: You have a large inventory of identical items. Y

19、ou want to predict how manyof these items will sell over the next 3 months.Problem 2: Youd like software to examine individual customer accounts, and for each account decide if it has been hacked/compromised.Should you treat these as classification or as regressionproblems?Treat both as classificati

20、on problems.Treat problem 1 as a classification problem, problem 2 as a regression problem.Treat problem 1 as a regression problem, problem 2 as a classification problem.Treat both as regression problems.Andrew NgAndrew NgIntroductionUnsupervisedLearningMachine LearningSupervised Learningx2x1Andrew

21、NgUnsupervised Learningx2x1Andrew NgAndrew NgAndrew NgIndividualsSource: Daphne KollerAndrew NgGenesIndividualsSource: Daphne KollerAndrew NgGenesOrganize computing clustersMarket segmentationSocial network analysisImage credit: NASA/JPL-Caltech/E. Churchwell (Univ. of Wisconsin, Madison)Astronomical data analysisAndrew NgCocktail party problemSpeaker #1Microphone #1Speaker #2Microphone #2Andrew NgMicrophone #1:Output #1:Microphone #2:Output #2:Microphone #1:Output #1:Microphone #2:Output